It may be useful to separate out the descriptive facts about non-cognitive aspects of education (they matter and they are malleable, even in very low-income settings) from the policy implications of these facts. One problem is that the results in this post are based on small scale studies, with dedicated and motivated teams. The results of an evaluation of the large early childhood scheme in the U.S., Head Start, is a mixed bag (http://www.acf.hhs.gov/programs/opre/hs/impact_study/reports/impact_study/executive_summary_final.pdf). So, I am not sure whether we have a proven to-scale program that improves non-cognitive outcomes among kids (I am not an expert, so would love to hear back on this). On what we should measure, I have been deeply influenced by Ed Lazear's analysis of the problem of high stakes testing (http://www.mitpressjournals.org/doi/abs/10.1162/qjec.121.3.1029?journalCode=qjec). Lazear's point is that when the cost of learning is high and the cost of monitoring is also high, its best to test with high stakes--announcing the test concepts beforehand. The intuition is simple and brilliant--if the cost of learning is super high, then the incentives from low-stakes testing are too low to result in any improvements. With high-stakes testing, the kids will focus on the concepts that are to be tested, but this may increase learning, because at least those concepts will be learnt. To put it another way, suppose I want to test tomorrow. I can say "I am going to test you with 8 questions, 2 each on addition, subtraction, division and multiplication" and you will get a dollar for every question you do right". Or, I can say "I am going to test you on 8 single-digit addition questions" and give you a dollar for every question you get right. Given that as a student, I have one night to study, in the first scenario, I may well decide not to study at all if I can learn only one concept, since the maximum I can get is $2 and the cost of studying is (say) $4. But in the second (high-stakes) test, I will study single-digit addition, because I stand to gain up to $8. So at least I will end up learning something in the second case. A contrived example---the paper is much nicer. From the comments, what I hear is that this is going to be costly to measure and potentially costly to change. Which is disappointing, since if you believe Lazear's line of reasoning, it suggests that this is something that school systems should not invest in currently. Not because non-cognitive aspects of education are unimportant, but because it will be very hard to provide appropriate incentives to improve them in a scaleable manner. Having said that, if you have 12 minutes, see this video, which argues that a change in the focus of schooling is critical: http://www.youtube.com/watch?v=mCbdS4hSa0s & thanks for the many enlightening comments.